The majority of cancer presents as a complex phenotype. A complex phenotype is manifest through gene-gene, and/or gene- environment interactions. Primary hepatocellular carcinoma (HCC) is an ideal paradigm for the investigation of complex cancer phenotypes in humans. Molecular studies of genetic alterations in tumors have identified p53 as a tumor suppressor gene commonly altered in HCC. Epidemiologic studies have firmly established the role of chronic hepatitis B virus infection (HBV) and aflatoxin B1 (AFB1) exposure as environmental risk factors. However, the majority of individuals exposed to HBV and AFB1 do not develop HCC. Genetic analysis is being used to assess the role of genes in well described pathways in determining disease. This approach merges gene mapping and candidate locus studies by including as candidates all the members of a pathway. Each gene of interest is tagged with multiple polymorphic sites in or near it. This approach has been applied to identify genetic factors modulating the risk of developing HCC among populations exposed to the hepatocarcinogen aflatoxin B1 (AFB1). The individual members of each family (GSTA1, GSTM1, GSTM3, GSTP, GSTT1, GST12, EPHX1, EPHX2) have been tagged with new or published polymorphisms. To assess their role in HCC risk, they have been examined in a nested case-control population. The loci GSTM1, GSTP, GSTT1, EPHX1 showed significant association with HCC risk while the EPHX2 locus was associated with age of onset. When results were stratified by the HBV status of the case, GSTM1 and GSTT1 were associated only in the HBV(+) cases, while GSTP was associated in the HBV(-) cases. These results indicate that these genes are candidates for more detailed functional and genetic analysis. Genetic information important in complex trait analysis may be accessible from the joint study of heritable variation and somatic tissue (tumor) variation in cancer. HCC tumor/normal pairs were examined using a collection genome-wide simple tandem repeat polymorphism markers and candidate loci. Allele loss patterns are observed to be complex. To assess whether underlying genetic patterns existed within the data, evolutionary tree building algorithms were used to examine the data. The branches of the tree had systematic, non-overlapping differences in the location of allele loss and the different branches also were observed to have significantly different rates of genome-wide allele loss rates. Finally, different candidate locus risk allele distributions for EPHX1 were observed among the different tree branches, suggesting that inclusion of such information may be important for gene discovery. HCC tumors have also been examined for mutations in tumor suppressor genes. Using RT-PCR we have begun to search for differences in expression and product in the tumor suppressor genes TSG1 and FHIT.